mirrors/tinygrad
2
0
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-06-24 02:14:17 +00:00
Code Issues Projects Releases Packages Wiki Activity
4,495 commits 280 branches 15 tags 325 MiB
Commit graph

4,495 commits

This branch
This branch
All branches
Author SHA1 Message Date
wozeparrot
6020595eb0
 		more tensor.py docs (#4686) 
wow much docs
 2024-05-22 21:28:26 +00:00
Francis Lam
721f9f6acf
 		test/external/verify_kernel: fix LOGKERNS variable name in comments (#4685) 
should've been changed with the LOGKERN to LOGKERNS change
 2024-05-22 17:08:40 -04:00
chenyu
f8f97562e0
 		remove File Specific Variables from env_vars.md (#4684) 2024-05-22 17:00:14 -04:00
chenyu
225dcab3be
 		prepend _ to broadcast_shape and deepwalk (#4683) 
* prepend `_` to broadcast_shape and deepwalk

internal only

* that too
 2024-05-22 16:39:05 -04:00
qazal
c5f5755328
 		correctness test for multireduce nested locals (#4682) 
* nested locals test

* move st
 2024-05-22 19:35:35 +03:00
chenyu
bc9be39dec
 		set timeout in search _try_compile_linearized_w_idx (#4677) 2024-05-22 12:30:31 -04:00
qazal
d12d412e8b
 		revert uops dtype in pattern matcher (#4681) 
This reverts commit 5f84cbb5df.
 2024-05-22 14:45:51 +03:00
Elias Wahl
acc0039cfc
 		Resume fix + scheduler for non weight decay params (#4679) 
* move ckpt dir

* fix resume. Add scheduler group
 2024-05-21 19:38:13 -04:00
chenyu
0f21aa0416
 		example kernel that triggers Memory access fault for resnet on red (#4678) 2024-05-21 18:59:36 -04:00
qazal
5f84cbb5df
 		keep UOps.CAST in PHI-GEP fold for unmatching dtypes (#4674) 
* these should be val.dtype

* cast float4 and float2 to root

* document tests

* 2 args

* fix assert

* match dtype

* no extra lines

* better fix
 2024-05-21 14:59:49 -04:00
qazal
458a3961eb
 		catch compile errors in uops tests (#4672) 
* use helper and compile

* llama beam=2

* ast length

* skip float4, fix hsa

* use empty tensors
 2024-05-21 12:20:35 +03:00
wozeparrot
00432496d7
 		feat: tinyboxgreen (#4366) 
* feat: tinyboxgreen

* feat: tinyboxgreenv2

* fix symlink weights

* fix: remove llama 2 70b for now

* feat: naming

* fix: remove extra cifar steps

* feat: disable mixtral on nvidia
 2024-05-20 22:39:34 -04:00
Timmy
de733d73cf
 		Multireduce Linearizer Tests (#4665) 
* updated tests

* make sure the upcasting tests actually causes the problem

* diff cleanup

* use UOpGraph utils

---------

Co-authored-by: qazal <qazal.software@gmail.com>
 2024-05-21 02:43:25 +03:00
chenyu
5e3fbbb33e
 		llama3 example add manual seed and log seed (#4667) 2024-05-20 19:09:57 -04:00
chenyu
8c99cc17f5
 		remove link to old adding_new_accelerators.md (#4666) 
fix #4657
 2024-05-20 19:05:23 -04:00
chenyu
c4089d169f
 		update BEAM_LOCAL_MAX to 1024 (#4664) 
we used 1024 for mlperf submission and result steps time is 20% faster. the default should not be worse
 2024-05-20 18:06:32 -04:00
chenyu
704cb1d8a0
 		fix conversation.py quantize (#4663) 
it used to be true for int8, not it's a string for int8 or nf4
 2024-05-20 17:36:37 -04:00
chenyu
ae861325ce
 		update llama sample for mac 32 input buffer limit (#4662) 
set default sampling params to function call to 0, and top k in llama3 to 25.
 2024-05-20 17:23:39 -04:00
Elias Wahl
993091adfa
 		loss scaler + nan fixes (#4661) 2024-05-20 17:08:35 -04:00
qazal
b33c827aed
 		UOps.RANGE toposort spec (#4660) 
* use iterator

* nested loops and outer loads

* uop after phi
 2024-05-20 23:38:20 +03:00
qazal
0d9e623d83
 		consolidate uops tests (#4659) 
* merge uoptimize

* move tests

* fix skip message
 2024-05-20 21:42:31 +03:00
Szymon Ożóg
1e7b7b2c3c
 		Fix flop coutning for mulacc (#4640) 
* Fix flop coutning for mulacc

* add test_simple_mulacc

* Update test_uops_stats.py

* Update test_uops_stats.py

* revert test_mulacc

* Test for MULACC vs MUL+ADD
 2024-05-20 12:06:00 -04:00
wozeparrot
b144d4b460
 		new llama3 example (#4576) 2024-05-19 22:42:23 -07:00
nimlgen
c9f7f2da70
 		nv hcq bind api (#4629) 
* hcq bind api for nv

* linter

* linter

* add test

* small comment
 2024-05-19 23:17:10 +03:00
qazal
d308f4fa9a
 		correctly insert UOps.END* in fuzz result (#4653) 2024-05-19 21:10:28 +03:00
chenyu
456aa0b656
 		update test_search kernel count (#4652) 
integration test that beaming 1 kernel increments kernel count by 1, and moved exiting test_kernel_count to TestTimeLinearizer
 2024-05-19 13:54:52 -04:00
qazal
954718e6bf
 		reorder DEFINE_GLOBAL in fuzz_uops (#4651) 
* globals base

* test: opt out of DEFINE_GLOBAL

* do it like ExecItem
 2024-05-19 20:51:31 +03:00
Léo
967e35f8b8
 		fix(beam): GlobalCounters kernel count increasing when clearing l2 (#4598) 
* fix(beam): GlobalCounters kernel count increasing when clearing l2

* fix: removed the NOSTATS var by adding do_update_stats to Tensor.realize()

* test(search): regression test for _time_program, should not increment kernel_count

* fix(test_search): unused var and now properly checking when l2 is cleared

* fix(test_search): added assert message

* fix(test_search): now testing public beam api for kcount

* ruff fixes

---------

Co-authored-by: Léo Paillé <leo.paille@enseirb-matmeca.fr>
 2024-05-19 10:03:47 -07:00
George Hotz
4753283221
 		LOOP -> RANGE (#4650) 2024-05-19 06:40:20 -07:00
chenyu
286b4dbdf2
 		compile raise CompileError and skip only RuntimeError in multiprocess… (#4646) 
* compile raise CompileError and skip only RuntimeError in multiprocess beam

renderer error with multiprocess should not be skipped by beam

* use `==` for dtype to dtype comparison

* that needs to be is

* typo
 2024-05-19 00:25:25 -04:00
chenyu
8a0d1ca7bb
 		CI test timeout 20 min -> 10 min (#4645) 
if it takes more than 10 usually setup fails anyway. also updated matmul_kfd -> matmul_amd in benchmark
 2024-05-18 13:58:28 -04:00
qazal
b0cb02f719
 		uops fuzzing infra (#4641) 
* base with bfs

* find paths

* get last

* try blocks

* Revert "try blocks"

This reverts commit 25f8e3fe85.

* this should be simpler

* full exec

* support debug

* fix lint

* add todo

* copy in_degree
 2024-05-18 20:19:57 +03:00
qazal
bf8f855838
 		assert kernel counts in unsupported fusions (#4643) 
* replace with comments

* not relevant

* update comment

* custom exception maybe

* fix LoadOps.VIEW
 2024-05-18 20:14:37 +03:00
qazal
a5204fe89d
 		refactor UOps.CONST (#4639) 
* delete more

* nit: dont need assign

* can this be simpler

* use scalars

* always cast

* clang needs cast

* format
 2024-05-18 10:07:36 +03:00
qazal
d0a2d40df3
 		root cause fix for UOps.CONST bad args (#4638) 
* delete that

* real fix
 2024-05-18 09:15:25 +03:00
George Hotz
9b464e34ea
 		increase speed of uops (#4637) 
* increase speed of uops

* not equal

* minor speedup
 2024-05-17 21:04:39 -07:00
George Hotz
b74cc1d01a
 		uops cleanup (#4634) 
* def add cleanup

* minor speedup

* add back ptx speed

* a little faster

* merge that

* only linearize once for ptx

* two graph rewrites for ptx, bug?
 2024-05-17 20:02:38 -07:00
George Hotz
07b350a8f4
 		new uops is an actual graph (#4560) 
* new uops is an actual graph

* it's way slower

* simpler

* fix define acc

* render_loop unique

* ops test pass

* add pattern matcher back, there's bugs

* rewrite

* use priority queue

* recursive children

* fix tests

* fix tests with SINK

* fix abstractions

* fix assembly

* simpler

* link define_acc

* fix DEFINE_ACC placement

* type verify

* full cmp

* fix cmp

* ACCESS_ACC

* insert DEFINE_ACC

* fix PHI

* recursive rewrite

* fix many tests

* sum collapse

* more patterns

* correct change

* fold arange

* fix that lin test

* space

* big folding rule works

* close

* has more maxes, meh

* cached node replace

* set changed

* simplest folding yet

* works

* works

* DIV

* all tests pass

* del

* fuzz linearizer fails

* sum_collapse

* test depth 2 cf

* fix lin test 14

* fix clang depth

* disable that

* failure 14 is fixed

* fix ptx

* failure 27 is fixed

* fix llama

* run_cnt

* Revert "Optimize PTX gated loads index calculation (#4304)"

This reverts commit d97d5a7689.

* fix uops loop

* fix ptx bugs

* add barrier

* print

* mem_type in ptx direct

* bypass tests that fail in CI but pass locally

* ptx remove ptr_ar

* more ptx passing

* fix ptx tests

* assert compile support

* remove  model inference benchmark from red
 2024-05-17 18:00:18 -07:00
nimlgen
daf57af3eb
 		move tc to renderers (#4631) 
* move tc to renderers

* missed import

* fix typo

* fix

* fix imports

* remove from tests

* fix 4607

* nv emulate timestamp

* time is int

* correct time
 2024-05-18 00:36:29 +03:00
chenyu
d70988dddf
 		add blob and raw=true for image in docs showcase (#4632) 
this should  render the image correctly
 2024-05-17 16:57:15 -04:00
nimlgen
10cf8e459b
 		hcq update queue in place (#4626) 
* do not self wait in hcq

* faster enqueue

* comments

* tests

* linter

* fix typo
 2024-05-17 22:18:20 +03:00
chenyu
ca1df20fa9
 		benchmark name fix - resnet eval is on eval data (#4628) 2024-05-17 12:56:12 -04:00
chenyu
c86adabe15
 		time with real global buffers in search (#4621) 
* filter fake buffers in search

* test that

* update test
 2024-05-17 12:36:23 -04:00
chenyu
e5d4e6a8aa
 		BEAM=2 in green CI for 100 TFLOPS (#4624) 2024-05-16 23:28:28 -04:00
chenyu
b3dd885ffb
 		cleanup double import from tinygrad.device in tensor.py (#4620) 2024-05-16 14:21:22 -04:00
uuuvn
639ea5b0f2
 		Metal linearizer failure 22 is flaky not just on CI (#4617) 
* METAL doesn't fail anymore, not just on CI

* oops
 2024-05-16 11:31:23 -04:00
qazal
f3f2b96583
 		pick schedule tests from external_test_opt (#4615) 
* conv tests

* misc

* that shouldnt const fold
 2024-05-16 15:43:41 +03:00
qazal
13200c6894
 		check simple_pads in all views (#4614) 2024-05-16 14:34:39 +03:00
qazal
0b464df605
 		base change scheduling spec (#4613) 
* spec and kernel cnt

* dont use half

* skip half
 2024-05-16 13:30:49 +03:00
nimlgen
65f7e3b3ab
 		nv setup constbuf4 (#4511) 
* nv correct constbuf 4

* compare results to cuda

* test fixed

* failed kernel

* repro

* revert this change
 2024-05-16 10:42:35 +03:00